This book summarizes techniques of fault prediction, detection, and identification, all included specifically in the data-driven fault diagnosis requirements within industrial processes, drawing from the combination of data science, machine learning, and domain-specific expertise. In the modern industrial processes, where efficiency, productivity, and safety stand as paramount pillars, the pursuit of fault diagnosis has become more crucial than ever. The widespread use of computer systems, along with new sensor hardware, generates significant quantities of real-time process data. It has been frequently asked what could be done with both the real-time and archived historical data, to not only promising efficiency but providing prospect of a brighter, more resilient future. This book starts with the definition, related work, and open test-bed for industrial process fault diagnosis. Then, it presents several data-driven methods on fault prediction (Part I), fault detection (Part II), and fault diagnosis (Part III), with consideration of properties of industrial processes, such as varying operation modes, non-Gaussian, nonlinearity. It distills cutting-edge methodologies and insights which may inspire for industrial practitioners, researchers, and academicians alike.